- Add persistent volume mounts for Go/npm caches (faster rebuilds)
- Add shell config with helpful aliases and custom prompt
- Add comprehensive devcontainer documentation
- Add pre-commit hooks for Go formatting and linting
- Use go-version-file in CI workflows instead of hardcoded versions
- Simplify docker compose commands with --wait flag
- Add gitignore entries for devcontainer auth files
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The LLM was confusing VMIDs because they weren't included in the
context. Now the formatted context shows:
### Container: ollama (VMID 200) on minipc
This prevents the AI from referencing the wrong VMID when generating
findings and recommendations.
100 samples was causing 326k+ input tokens which is expensive.
24 samples (hourly resolution) still provides good pattern visibility
while significantly reducing token cost.
Estimated reduction: ~75% fewer metric tokens.
Modern LLMs have 100k+ token contexts. 100 samples over 24h gives
~15 minute resolution while adding minimal token overhead.
This lets the LLM see fine-grained patterns, short spikes, and
accurately distinguish anomalies from normal behavior.
The in-memory MetricsHistory only retains 24 hours of data, not 7 days.
Changed computeGuestMetricSamples to use trendWindow24h instead of
trendWindow7d, and reduced sample count from 24 to 12 points.
This ensures the LLM actually receives metric samples in the context,
which wasn't happening before because the 7-day query returned empty data.
Instead of relying on pre-computed trend heuristics (which can be misleading
for edge cases like step changes vs continuous growth), we now pass downsampled
raw data points to the LLM so it can interpret patterns directly.
Changes:
- Add MetricSamples field to ResourceContext
- Add DownsampleMetrics() to reduce data points for LLM consumption
- Add formatMetricSamples() to format data compactly (e.g., 'Disk: 26→26→31%')
- Add computeGuestMetricSamples() to gather 7-day sampled history
- Populate MetricSamples for VMs and containers during context build
- Add History section to formatted context output
The LLM now sees actual patterns like 'stable for 6 days then jumped' rather
than just '45.8%/day growth rate' - allowing for much more nuanced interpretation.
This approach:
- Leverages LLM's pattern recognition instead of hard-coded heuristics
- Provides 7 days of data (~24 samples) for context on normal behavior
- Uses minimal tokens due to compact formatting with deduplication
- Is more future-proof as LLMs improve
Example output:
**History (7d sampled, oldest→newest)**: Disk: 26→26→26→26→26→31%
Refs: Frigate disk usage false positive investigation
- Frontend: Add ociImage memo to extract clean image name from osTemplate
- Frontend: Show OCI image name in type badge tooltip
- Frontend: Display OCI image in OS column when no guest agent info available
- Frontend: Include ociImage in AI context data for selected OCI containers
- Backend: Differentiate OCI containers as 'oci_container' type in AI context
- Backend: Add Metadata field to ResourceContext for extensibility
- Backend: Include oci_image in container metadata for AI analysis
- Backend: Update section heading to 'LXC/OCI Containers' in AI context
This follows Docker container patterns to avoid duplicating work.
Phase 2 of Pulse AI differentiation:
- Create internal/ai/baseline package for learned baselines
- Implement statistical baseline learning with mean, stddev, percentiles
- Add z-score based anomaly detection with severity classification
(low, medium, high, critical based on standard deviations)
- Integrate baseline provider into context builder
- Wire baseline store into patrol service with adapters
- Add anomaly enrichment to resource contexts
Key features:
- Learn computes baseline from historical metric data points
- IsAnomaly and CheckAnomaly detect deviations from normal
- Persists baselines to disk as JSON for durability
- Formatted anomaly descriptions for AI consumption
Example: 'Memory is high above normal (85.2% vs typical 42.1% ± 8.3%)'
The baseline store needs to be initialized and triggered to learn
from metrics history. Next step is adding the learning loop.
All tests passing.
Phase 1 of Pulse AI differentiation:
- Create internal/ai/context package with types, trends, builder, formatter
- Implement linear regression for trend computation (growing/declining/stable/volatile)
- Add storage capacity predictions (predicts days until 90% and 100%)
- Wire MetricsHistory from monitor to patrol service
- Update patrol to use buildEnrichedContext instead of basic summary
- Update patrol prompt to reference trend indicators and predictions
This gives the AI awareness of historical patterns, enabling it to:
- Identify resources with concerning growth rates
- Predict capacity exhaustion before it happens
- Distinguish between stable high usage vs growing problems
- Provide more actionable, time-aware insights
All tests passing. Falls back to basic summary if metrics history unavailable.
